DeepQLearning_Pytorch_SpaceInvaders.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "DeepQLearning-Pytorch-SpaceInvaders.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/rubencg195/Pytorch-Tutorials/blob/master/DeepQLearning_Pytorch_SpaceInvaders.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "metadata": {
        "id": "zK9yKwBcMMQY",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Imports\n",
        "\n",
        "https://github.com/simoninithomas/Deep_reinforcement_learning_Course/blob/master/Deep%20Q%20Learning/Space%20Invaders/DQN%20Atari%20Space%20Invaders.ipynb"
      ]
    },
    {
      "metadata": {
        "id": "1IFW5W87weYB",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "%matplotlib inline\n",
        "\n",
        "import os\n",
        "import time\n",
        "import gym\n",
        "import math\n",
        "import random\n",
        "import numpy as np"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "xhc6iOZzYdSK",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 222
        },
        "outputId": "af8fbae6-f96e-445b-d323-b57c6e56a8ad"
      },
      "cell_type": "code",
      "source": [
        "OS_Windows = False\n",
        "\n",
        "if OS_Windows:\n",
        "    !ECHO %CurrentDir%\n",
        "    drive_path = \"./\"\n",
        "else:\n",
        "    #Install Google Drive\n",
        "    from google.colab import drive\n",
        "    drive.mount('/content/gdrive')\n",
        "    drive_path = \"gdrive/My Drive/Colab Notebooks/\"\n",
        "    !ls  \"gdrive/My Drive/Colab Notebooks/\"\n"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/gdrive; to attempt to forcibly remount, call drive.mount(\"/content/gdrive\", force_remount=True).\n",
            " Beginner\t\t\t\t\t    KRAGGLE\n",
            " Books\t\t\t\t\t\t    log_deepq_pytorch.txt\n",
            "'Convolutionary Network Tutorial - Pytorch.ipynb'   model_deepq_pytorch.ckpt\n",
            "'Copy of DeepQLearning-Pytorch.ipynb'\t\t    MUN\n",
            "'DATA LOADING AND PROCESSING TUTORIAL.ipynb'\t    PONG\n",
            " DeepQLearning-Pytorch-SpaceInvaders.ipynb\t    PyTorch\n",
            "'DeepQ Learning Tutorial.ipynb'\t\t\t    STATISTICS\n",
            "'DEEPQ NETWORK SPACE INVADERS.ipynb'\t\t    Theory\n",
            " Intermidiate\t\t\t\t\t    Udacity\n",
            "'Keras Tutorial'\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "DKsS_afGwtJL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# http://pytorch.org/\n",
        "from os.path import exists\n",
        "from wheel.pep425tags import get_abbr_impl, get_impl_ver, get_abi_tag\n",
        "platform = '{}{}-{}'.format(get_abbr_impl(), get_impl_ver(), get_abi_tag())\n",
        "cuda_output = !ldconfig -p|grep cudart.so|sed -e 's/.*\\.\\([0-9]*\\)\\.\\([0-9]*\\)$/cu\\1\\2/'\n",
        "accelerator = cuda_output[0] if exists('/dev/nvidia0') else 'cpu'\n",
        "\n",
        "!pip install -q http://download.pytorch.org/whl/{accelerator}/torch-0.4.1-{platform}-linux_x86_64.whl torchvision"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "QgxjOBwBwvCs",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "import torch as T\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.nn.functional as F\n",
        "#import torchvision.transforms as T\n",
        "from torch.autograd import Variable"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "p-r-64kjMReY",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Hyperparameters"
      ]
    },
    {
      "metadata": {
        "id": "luCNGmE-JZnh",
        "colab_type": "code",
        "cellView": "both",
        "outputId": "650c5b2f-8741-44b3-d72c-5e6276913d87",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 168
        }
      },
      "cell_type": "code",
      "source": [
        "#Environment\n",
        "env_name = 'SpaceInvaders-v0' #@param [\"DoomBasic-v0\", \"Breakout-v0\", \"Pong-v0\"] {allow-input: true}\n",
        "\n",
        "env = gym.make(env_name)\n",
        "\n",
        "print(\"Action Space: \", env.action_space.n)\n",
        "print(\"Action Space: \", env.observation_space.shape)\n",
        "\n",
        "possible_actions = np.array(np.identity(env.action_space.n,dtype=int).tolist()) \n",
        "print(\"Possible Actions: \\n\", possible_actions)\n",
        "\n",
        "\n",
        "### MODEL HYPERPARAMETERS\n",
        "frame_width  = 84                #@param {type:\"number\"} \n",
        "frame_height = 110               #@param {type:\"number\"} \n",
        "\n",
        "### PREPROCESSING HYPERPARAMETERS\n",
        "# Number of frames stacked\n",
        "stack_size = 4                   #@param {type:\"number\"}              \n",
        "state_size = [frame_height, frame_width, stack_size]        # Our input is a stack of 4 frames hence 110x84x4 (Width, height, channels)\n",
        "\n",
        "# 6 possible actions\n",
        "##action_size = env.action_space.n   #@param {type:\"raw\"}\n",
        "action_size = 6   #@param {type:\"raw\"}\n",
        "\n",
        "# Alpha (aka learning rate)\n",
        "learning_rate =  0.00025         #@param {type:\"number\"}      \n",
        "\n",
        "### TRAINING HYPERPARAMETERS\n",
        "\n",
        "# Total episodes for training\n",
        "total_episodes = 20000              #@param {type:\"number\"}\n",
        "# Max possible steps in an episode\n",
        "max_steps = 50000                #@param {type:\"number\"} \n",
        "# Batch size\n",
        "batch_size = 10                #@param {type:\"number\"}              \n",
        "\n",
        "# Exploration parameters for epsilon greedy strategy\n",
        "\n",
        "# exploration probability at start\n",
        "explore_start = 1.0              #@param {type:\"number\"}\n",
        "# minimum exploration probability \n",
        "explore_stop = 0.01              #@param {type:\"number\"}\n",
        "# exponential decay rate for exploration prob\n",
        "decay_rate = 0.000005             #@param {type:\"number\"}         \n",
        "\n",
        "# Q learning hyperparameters\n",
        "# Discounting rate\n",
        "gamma = 0.9                      #@param {type:\"number\"}                \n",
        "\n",
        "### MEMORY HYPERPARAMETERS\n",
        "\n",
        "# Number of experiences stored in the Memory when initialized for the first time\n",
        "pretrain_length = batch_size  \n",
        "# Number of experiences the Memory can keep\n",
        "memory_size = 50000            #@param {type:\"number\"}       \n",
        "\n",
        "\n",
        "### MODIFY THIS TO FALSE IF YOU JUST WANT TO SEE THE TRAINED AGENT\n",
        "training = True                 #@param [\"False\", \"True\"] {type:\"raw\"}\n",
        "\n",
        "## TURN THIS TO TRUE IF YOU WANT TO RENDER THE ENVIRONMENT\n",
        "episode_render = False           #@param [\"False\", \"True\"] {type:\"raw\"}\n",
        "\n",
        "# Here we create an hot encoded version of our actions\n",
        "# possible_actions = [[1, 0, 0, 0, 0, 0, 0, 0], [0, 1, 0, 0, 0, 0, 0, 0, 0]...]\n",
        "\n",
        "# We stack 4 frames\n",
        "stack_size = 4  #@param {type:\"number\"}       \n",
        "\n",
        "target_update = 50 #@param {type:\"number\"}       \n",
        "\n",
        "load_weigths  = True            #@param [\"False\", \"True\"] {type:\"raw\"}"
      ],
      "execution_count": 31,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Action Space:  6\n",
            "Action Space:  (210, 160, 3)\n",
            "Possible Actions: \n",
            " [[1 0 0 0 0 0]\n",
            " [0 1 0 0 0 0]\n",
            " [0 0 1 0 0 0]\n",
            " [0 0 0 1 0 0]\n",
            " [0 0 0 0 1 0]\n",
            " [0 0 0 0 0 1]]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "cuqUsfeF28vm",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Preprocessing"
      ]
    },
    {
      "metadata": {
        "id": "9PUHiC9jO_Fu",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "**Stack frames**\n",
        "\n",
        "As explained in this really good article we stack frames.\n",
        "\n",
        "https://danieltakeshi.github.io/2016/11/25/frame-skipping-and-preprocessing-for-deep-q-networks-on-atari-2600-games/\n",
        "\n",
        "Stacking frames is really important because it helps us to give have a sense of motion to our Neural Network.\n",
        "\n",
        "BUT, we don't stack each frames, we skip 4 frames at each timestep. This means that only every fourth frame is considered. And then, we use this frame to form the stack_frame.\n",
        "\n",
        "The frame skipping method is already implemented in the library.\n",
        "\n",
        "First we preprocess frame\n",
        "Then we append the frame to the deque that automatically removes the oldest frame\n",
        "Finally we build the stacked state\n",
        "This is how work stack:\n",
        "\n",
        "For the first frame, we feed 4 frames\n",
        "At each timestep, we add the new frame to deque and then we stack them to form a new stacked frame\n",
        "And so on stack\n",
        "If we're done, we create a new stack with 4 new frames (because we are in a new episode)."
      ]
    },
    {
      "metadata": {
        "id": "IhmIM3ty2-R-",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from skimage import transform      # Help us to preprocess the frames\n",
        "from skimage.color import rgb2gray # Help us to gray our frames\n",
        "import matplotlib.pyplot as plt    # Display graphs\n",
        "import warnings                    # This ignore all the warning messages that are normally printed during the training because of skiimage\n",
        "warnings.filterwarnings('ignore')\n",
        "\n",
        "def preprocess_frame(frame, display=False):\n",
        "  \"\"\"\n",
        "  Reduce complexity\n",
        "  Steps:\n",
        "  1. Grayscale\n",
        "  2. Crop\n",
        "  3. Normalize \n",
        "  4. Resize\n",
        "  \"\"\" \n",
        "  gray       = rgb2gray(frame)\n",
        "  cropped    = gray[8:-12, 4:-12] \n",
        "  normalized = cropped/255.0 \n",
        "  resized    = transform.resize(normalized, [110,84])\n",
        "  \n",
        "  if display: \n",
        "\n",
        "    fig, ax = plt.subplots(nrows=1, ncols=5, sharey=True, sharex=True, figsize=(25,50))\n",
        "\n",
        "    plt.subplot(1, 5, 1)\n",
        "    plt.imshow(frame)\n",
        "    \n",
        "    plt.subplot(1, 5, 2)\n",
        "    plt.imshow(gray)\n",
        "    \n",
        "    plt.subplot(1, 5, 3)\n",
        "    plt.imshow(cropped)\n",
        "    \n",
        "    plt.subplot(1, 5, 4)\n",
        "    plt.imshow(normalized)\n",
        "    \n",
        "        \n",
        "    plt.subplot(1, 5, 5)\n",
        "    plt.imshow(resized)\n",
        "\n",
        "    plt.show()\n",
        "    \n",
        "    #print(\"New  Shape: \", resized.shape)\n",
        "\n",
        "  return resized # 110x84x1 frame\n",
        "  \n",
        "  \n",
        "  "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "T_smLEsFO3qI",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "def stack_frames(stacked_frames, state, is_new_episode):\n",
        "    # Preprocess frame\n",
        "    frame = preprocess_frame(state)\n",
        "    \n",
        "    if is_new_episode:\n",
        "        # Clear our stacked_frames\n",
        "        stacked_frames = deque([np.zeros((110,84), dtype=np.int) for i in range(stack_size)], maxlen=4)\n",
        "        \n",
        "        # Because we're in a new episode, copy the same frame 4x\n",
        "        stacked_frames.append(frame)\n",
        "        stacked_frames.append(frame)\n",
        "        stacked_frames.append(frame)\n",
        "        stacked_frames.append(frame)\n",
        "        \n",
        "        # Stack the frames\n",
        "        stacked_state = np.stack(stacked_frames, axis=2)\n",
        "        \n",
        "    else:\n",
        "        # Append frame to deque, automatically removes the oldest frame\n",
        "        stacked_frames.append(frame)\n",
        "        \n",
        "        # Build the stacked state (first dimension specifies different frames)\n",
        "        stacked_state = np.stack(stacked_frames, axis=2) \n",
        "\n",
        "        #print(\"stacked_frames shape\", len(stacked_frames), stacked_frames[0].shape )\n",
        "        #print(\"stacked_state shape\", stacked_state.shape)\n",
        "        #Output \n",
        "        #stacked_frames shape 4 (110, 84)\n",
        "        #stacked_state shape (110, 84, 4)\n",
        "    \n",
        "    return stacked_state, stacked_frames"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "WjcvjrB6gqt_",
        "colab_type": "code",
        "outputId": "95656090-56f4-4ac9-81fc-084db4bf2356",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 623
        }
      },
      "cell_type": "code",
      "source": [
        "obs = env.reset()\n",
        "print(\"Shape\",  obs.shape )\n",
        "preprocess_frame(obs, display=True)"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Shape (210, 160, 3)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABZkAAAFfCAYAAAAszcLlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nzJIkSZIkSZKkQZHRDgAAAAdJREFUof0/uswKQTO5fjgAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f4ba17787b8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "array([[0.        , 0.        , 0.        , ..., 0.        , 0.        ,\n",
              "        0.        ],\n",
              "       [0.00151918, 0.00151918, 0.00151918, ..., 0.00147554, 0.        ,\n",
              "        0.        ],\n",
              "       [0.00136726, 0.00136726, 0.00029298, ..., 0.00132799, 0.        ,\n",
              "        0.        ],\n",
              "       ...,\n",
              "       [0.        , 0.        , 0.        , ..., 0.        , 0.        ,\n",
              "        0.        ],\n",
              "       [0.        , 0.        , 0.        , ..., 0.00115   , 0.00115   ,\n",
              "        0.00115   ],\n",
              "       [0.000805  , 0.000805  , 0.00116643, ..., 0.001265  , 0.001265  ,\n",
              "        0.001265  ]])"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "metadata": {
        "id": "2eQ0MQlhw7JO",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Deep Q Network"
      ]
    },
    {
      "metadata": {
        "id": "iOGL5AYs1vWz",
        "colab_type": "code",
        "outputId": "885d1f59-36ea-4327-d502-800f9fe3e1d0",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 67
        }
      },
      "cell_type": "code",
      "source": [
        "def outputSize(in_size, kernel_size, stride, padding):\n",
        "    output = int((in_size - kernel_size + 2*(padding)) / stride) + 1\n",
        "    return (output)\n",
        "\n",
        "#Input (3, 110, 84) Conv, Output specified = 18 for batch size. for Maxpool layer number 1 remains unchanged\n",
        "print(\"Convolution #1\", (32,\n",
        "       outputSize(in_size=110, kernel_size=8, stride=4, padding=1 ),\n",
        "       outputSize(in_size=84 , kernel_size=8, stride=4, padding=1 ),\n",
        "      ))\n",
        "\n",
        "#Input (32, 27, 20) Conv, Output specified = 18 for batch size. for Maxpool layer number 1 remains unchanged\n",
        "print(\"Convolution #2\", (64,\n",
        "       outputSize(in_size=27 , kernel_size=4, stride=2, padding=0 ),\n",
        "       outputSize(in_size=20 , kernel_size=4, stride=2, padding=0 ),\n",
        "      ))\n",
        "\n",
        "#Input (64, 12, 9) Conv, Output specified = 18 for batch size. for Maxpool layer number 1 remains unchanged\n",
        "print(\"Convolution #3\", (128,\n",
        "       outputSize(in_size=12, kernel_size=3, stride=2, padding=0 ),\n",
        "       outputSize(in_size=9 , kernel_size=3, stride=2, padding=0 ),\n",
        "      ))\n"
      ],
      "execution_count": 35,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Convolution #1 (32, 27, 20)\n",
            "Convolution #2 (64, 12, 9)\n",
            "Convolution #3 (128, 5, 4)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "tIXD2ymqw89u",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class DeepQ(nn.Module):\n",
        "  def __init__(self, learning_rate, num_actions, input_shape ):\n",
        "    super(DeepQ, self).__init__()\n",
        "    self.conv1  = nn.Conv2d(4,   32 , kernel_size=8, stride=4, padding=1 )\n",
        "    self.bn1    = nn.BatchNorm2d(32)\n",
        "    \n",
        "    self.conv2  = nn.Conv2d(32,   64 , kernel_size=4, stride=2, padding=0 )\n",
        "    self.bn2    = nn.BatchNorm2d(64)\n",
        "    \n",
        "    self.conv3  = nn.Conv2d(64,   128, kernel_size=3, stride=2, padding=0 )\n",
        "    self.bn3    = nn.BatchNorm2d(128)\n",
        "    \n",
        "    self.fc1    = nn.Linear(128 * 5 * 4  , 512)\n",
        "    self.fc2    = nn.Linear(512          , num_actions)\n",
        "    \n",
        "    self.optim  = optim.Adam(self.parameters(), lr=learning_rate)\n",
        "    self.loss   = nn.MSELoss()\n",
        "    self.device = T.device('cuda:0' if T.cuda.is_available() else 'cpu')\n",
        "    self.to(self.device)\n",
        "    \n",
        "  def forward(self, obs):\n",
        "    #print(\"Forward Input Shape\", obs.shape)\n",
        "    obs     =obs.to(self.device)\n",
        "    obs     = obs.view(-1, 4, 110, 84 )\n",
        "    obs     = F.relu(self.bn1(self.conv1(obs)))\n",
        "    obs     = F.relu(self.bn2(self.conv2(obs)))\n",
        "    obs     = F.relu(self.bn3(self.conv3(obs)))\n",
        "    obs     = obs.view(-1, 128 * 5 * 4 )\n",
        "    obs     = F.relu(self.fc1(obs))\n",
        "    \n",
        "    actions = self.fc2(obs) #No Relu, raw data represents Q(s, a) value\n",
        "    return actions\n",
        "    "
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "_LEF8qS-f5VL",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Memory"
      ]
    },
    {
      "metadata": {
        "id": "N1yQnioNFcDL",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "from collections import deque# Ordered collection with ends\n",
        "class Memory():\n",
        "  def __init__(self, max_size):\n",
        "    self.buffer = deque(maxlen = max_size)\n",
        "    \n",
        "  def add(self, experience):\n",
        "    self.buffer.append(experience)\n",
        "   \n",
        "  def len(self):\n",
        "    return len(self.buffer)\n",
        "  \n",
        "  def sample(self, batch_size):\n",
        "#     print('Buffer Size', len(self.buffer) )\n",
        "    return random.sample(self.buffer, batch_size)\n",
        "#     buffer_size = len(self.buffer)\n",
        "#     index       = np.random.choice(\n",
        "#         np.arange(buffer_size), \n",
        "#         size    = batch_size,\n",
        "#         replace = False \n",
        "#     )\n",
        "#     return [self.buffer[i] for i in index]"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "n718lFeM18xc",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Agent"
      ]
    },
    {
      "metadata": {
        "id": "chWxq6Ni17zx",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "class Agent(object):\n",
        "  def __init__(self, \n",
        "                  stack_size,\n",
        "                  state_size,\n",
        "                  action_size,\n",
        "                  learning_rate,\n",
        "                  total_episodes,\n",
        "                  max_steps,\n",
        "                  batch_size,\n",
        "                  explore_start,\n",
        "                  explore_stop,\n",
        "                  decay_rate,\n",
        "                  gamma,\n",
        "                  pretrain_length,\n",
        "                  memory_size,\n",
        "                  training ,\n",
        "                  episode_render,\n",
        "                  possible_actions,\n",
        "                  target_update\n",
        "              ):\n",
        "    \n",
        "    self.stack_size         = stack_size\n",
        "    self.state_size         = state_size\n",
        "    self.action_size        = action_size\n",
        "    self.learning_rate      = learning_rate\n",
        "    self.total_episodes     = total_episodes\n",
        "    self.max_steps          = max_steps\n",
        "    self.batch_size         = batch_size\n",
        "    self.explore_start      = explore_start\n",
        "    self.explore_stop       = explore_stop\n",
        "    self.decay_rate         = decay_rate\n",
        "    self.gamma              = gamma\n",
        "    self.pretrain_length    = pretrain_length\n",
        "    self.memory_size        = memory_size\n",
        "    self.training           = training\n",
        "    self.episode_render     = episode_render\n",
        "    self.possible_actions   = possible_actions\n",
        "    self.target_update      = target_update\n",
        "    self.Q_eval             = DeepQ(\n",
        "        learning_rate       =  self.gamma\n",
        "        ,num_actions        =  self.action_size\n",
        "        ,input_shape        =  self.state_size\n",
        "    )\n",
        "    self.Q_target           = DeepQ(\n",
        "        learning_rate       =  self.gamma,\n",
        "        num_actions         =  self.action_size,\n",
        "        input_shape         =  self.state_size\n",
        "    )  \n",
        "     \n",
        "    if load_weigths != None and load_weigths == True:\n",
        "      print(\"Loadinig Models\")      \n",
        "      self.Q_target.load_state_dict(\n",
        "          T.load(\n",
        "              drive_path+\"/model_deepq_pytorch.ckpt\", \n",
        "              map_location=self.Q_target.device\n",
        "          )\n",
        "      )  \n",
        "      self.Q_target.load_state_dict(\n",
        "          T.load(\n",
        "              drive_path+\"/model_deepq_pytorch.ckpt\", \n",
        "              map_location=self.Q_target.device\n",
        "          )\n",
        "      )\n",
        "    else:\n",
        "      print(\"Loading New Weights\")\n",
        "    \n",
        "    \n",
        "    device = T.device(\"cuda\" if T.cuda.is_available() else \"cpu\") # PyTorch v0.4.0\n",
        "    self.Q_eval.to(device)\n",
        "    self.Q_target.to(device)\n",
        "    \n",
        "\n",
        "    #Target Policy\n",
        "    self.memory = Memory(max_size = self.memory_size)\n",
        "    # Initialize deque with zero-images one array for each image\n",
        "    self.stacked_frames  =  deque(\n",
        "        [np.zeros((110,84), dtype=np.int) for i in range(stack_size)], \n",
        "        maxlen=4\n",
        "    )\n",
        "\n",
        "    \n",
        "    print(\n",
        "      '_'*10 +  'HYPERPARAMETERS'  +   '_'*10  + '\\n'     ,\n",
        "      '\\nstack_size        '      ,self.stack_size        , \n",
        "      '\\nstate_size        '      ,self.state_size        ,\n",
        "      '\\naction_size       '      ,self.action_size       ,\n",
        "      '\\nlearning_rate     '      ,self.learning_rate     ,\n",
        "      '\\ntotal_episodes    '      ,self.total_episodes    ,\n",
        "      '\\nmax_steps         '      ,self.max_steps         ,\n",
        "      '\\nbatch_size        '      ,self.batch_size        ,\n",
        "      '\\nexplore_start     '      ,self.explore_start     ,\n",
        "      '\\nexplore_stop      '      ,self.explore_stop      ,\n",
        "      '\\ndecay_rate        '      ,self.decay_rate        ,\n",
        "      '\\ngamma             '      ,self.gamma             ,\n",
        "      '\\npretrain_length   '      ,self.pretrain_length   ,\n",
        "      '\\nmemory_size       '      ,self.memory_size       ,\n",
        "      '\\ntraining          '      ,self.training          ,\n",
        "      '\\nepisode_render    '      ,self.episode_render    ,\n",
        "      '\\ntarget_update     '      ,self.target_update     ,\n",
        "      '\\npossible_actions\\n'      ,self.possible_actions  ,\n",
        "      '\\n'\n",
        "    )\n",
        "  \n",
        "  def save_log(self, episode,total_reward, explore_probability, loss, running_loss , time, step, exp, running, action ):\n",
        "    with open(drive_path+\"/log_deepq_pytorch.txt\" , \"a\") as myfile:\n",
        "      log = '{}\\tEpisode:\\t{}\\tTotalReward:\\t{}\\tExploreP:\\t{:.4f}\\tRunningLoss:\\t{:.4f}\\tLoss:\\t{:.4f}\\tTimeLapse:\\t{}\\tMin.\\tStep:\\t{}\\t{}\\tAction\\t{}'.format( \n",
        "          running, episode,total_reward, explore_probability, running_loss , loss , time, step, exp, action\n",
        "      ) \n",
        "      myfile.write(log)\n",
        "      print(log)\n",
        "      \n",
        "  def predict_action(self, decay_step, state):\n",
        "    #EPSILON GREEDY STRATEGY\n",
        "    #Choose action a from state s using epsilon greedy\n",
        "    #First we randomize a number\n",
        "    exp_exp_tradeoff       = np.random.rand()\n",
        "    explore_probability    = self.explore_stop         \\\n",
        "    + (self.explore_start  - self.explore_stop  )      \\\n",
        "    * np.exp(-decay_rate   * decay_step    )\n",
        "    \n",
        "    if(explore_probability > exp_exp_tradeoff):\n",
        "      #Make a random action (Exploration)\n",
        "      choice               = random.randint(1, len(self.possible_actions)) - 1\n",
        "      action               = self.possible_actions[choice]\n",
        "\n",
        "    else:\n",
        "      #Get action from QNet (Explotation)\n",
        "      action   =  self.Q_eval.forward(T.Tensor(state))\n",
        "      action   =  T.argmax(action).item()\n",
        "      \n",
        "    return action, explore_probability, (explore_probability > exp_exp_tradeoff)\n",
        "  \n",
        "  def train(self):\n",
        "    #Initialize the decay_rate (This will reduce epsilon)\n",
        "    self.decay_step        = 0\n",
        "    self.start_time        = time.time()\n",
        "    \n",
        "    self.state             = env.reset()\n",
        "    preprocess_frame(self.state, display=True)\n",
        "    \n",
        "    self.reward_list       = []\n",
        "    \n",
        "    for episode in range(self.total_episodes):\n",
        "      #Set Step to 0\n",
        "      self.step            = 0\n",
        "      \n",
        "\n",
        "      #Initialize the rewards of the episode\n",
        "      self.episode_rewards = []\n",
        "      self.episode_loss    = []\n",
        "      self.episode_actions = []\n",
        "      \n",
        "      #Make a new episode and observe \n",
        "      self.state                  = env.reset()\n",
        "      \n",
        "      #Remember that stack frame function also call our preprocess function\n",
        "      self.state, self.stacked_frames  = stack_frames(\n",
        "          self.stacked_frames, \n",
        "          self.state, \n",
        "          is_new_episode=True\n",
        "      )\n",
        "      \n",
        "      while self.step < self.max_steps:\n",
        "        self.step       += 1\n",
        "        \n",
        "        #Increase the decay step\n",
        "        self.decay_step += 1\n",
        "        \n",
        "        #Predict the action to take and take it\n",
        "        self.action, self.explore_probability, self.exploration_vs_explotation = self.predict_action(\n",
        "            self.decay_step,\n",
        "            self.state\n",
        "        )\n",
        "        \n",
        "        #Perform the action and get the next_state, reward, and done info\n",
        "        self.next_state, self.reward, self.done, _ = env.step(np.argmax(self.action))\n",
        "        self.episode_actions.append(np.argmax(self.action))\n",
        "        \n",
        "        if episode_render:\n",
        "          env.render()\n",
        "          \n",
        "        #Add the reward to the total reward\n",
        "        self.episode_rewards.append(self.reward)\n",
        "        \n",
        "        if self.done:\n",
        "          #The episode ends so no next state\n",
        "          self.next_state = np.zeros( (110, 84) , dtype=np.int )\n",
        "          \n",
        "          self.next_state, self.stacked_frames = stack_frames(\n",
        "              self.stacked_frames, \n",
        "              self.next_state,\n",
        "              is_new_episode=False\n",
        "          )\n",
        "          \n",
        "          #Set Step = max_steps to end the episode\n",
        "          self.step = self.max_steps\n",
        "          \n",
        "          #Get the total reward of the episode\n",
        "          total_reward = np.sum(self.episode_rewards)\n",
        "          total_loss   = np.sum(self.episode_loss)\n",
        "          \n",
        "          self.save_log(\n",
        "              episode,\n",
        "              total_reward,\n",
        "              self.explore_probability,\n",
        "              total_loss / self.step ,\n",
        "              total_loss , \n",
        "              ( time.time() - self.start_time) / 60,\n",
        "              self.step,\n",
        "              \"Exploration\"  if (self.exploration_vs_explotation) else \"Explotation\",\n",
        "              \"Done:\",\n",
        "              self.episode_actions\n",
        "          )\n",
        "          \n",
        "          self.reward_list.append( (episode, total_reward) )\n",
        "          \n",
        "          #Store transition <st,at,rt+1,st+1> in memoryD\n",
        "          self.memory.add( \n",
        "              (self.state, self.action, self.reward, self.next_state, self.done)\n",
        "          )\n",
        "          \n",
        "        else:\n",
        "          #Stack the frame of the next_state\n",
        "          self.next_state, stacked_frames = stack_frames(\n",
        "            self.stacked_frames, \n",
        "            self.next_state,\n",
        "            is_new_episode=False\n",
        "          )\n",
        "          \n",
        "          #Add experence to memory\n",
        "          self.memory.add( \n",
        "              (self.state, self.action, self.reward, self.next_state, self.done)\n",
        "          )\n",
        "          \n",
        "          #st+1 is now our current state\n",
        "          self.state = self.next_state\n",
        "          \n",
        "        ###LEARNING\n",
        "        #obtain random mini-batch from memory\n",
        "        if self.memory.len() > self.batch_size:\n",
        "          \n",
        "          batch              = self.memory.sample(self.batch_size)\n",
        "          states_mb          = T.FloatTensor(np.array([each[0] for each in batch])).to(self.Q_eval.device)\n",
        "          #actions_mb         = np.array([each[1] for each in batch])\n",
        "          rewards_mb         = T.FloatTensor(np.array([each[2] for each in batch])).to(self.Q_eval.device)\n",
        "          next_states_mb     = T.FloatTensor(np.array([each[3] for each in batch])).to(self.Q_eval.device)\n",
        "          dones_mb           = np.array([each[4] for each in batch])\n",
        "\n",
        "          target_Qs_batch    = []\n",
        "          \n",
        "          #Get Q values for next state\n",
        "          self.Qs_next_state       = self.Q_target.forward(next_states_mb)\n",
        "          self.Qs_state            = self.Q_eval.forward(states_mb)\n",
        "#           self.Qtarget             = T.zeros_like(self.Qs_next_state) \n",
        "          self.Qtarget             = self.Qs_state.clone() \n",
        "         \n",
        "          #Set Q_target = r if the episode ends at s+1, otherwise\n",
        "          #set Q_target = r + gamma*maxQ(s', a')\n",
        "          for i in range  (0, len(batch)):\n",
        "            terminal                    = dones_mb[i]\n",
        "            maxAction                   = T.argmax(self.Qs_next_state[i]).to(self.Q_eval.device)\n",
        "            #if we are in a terminal state only equals reward\n",
        "            if terminal:\n",
        "              #print(\"Row {} Col {} Row Val {}, Col Val {}\".format( i,maxAction, self.Qtarget[i,maxAction].item() , rewards_mb[i].item() ) )\n",
        "              self.Qtarget[i,maxAction] = rewards_mb[i]\n",
        "              \n",
        "            else:\n",
        "              #print(\"Row {} Col {} Row Val {}, Col Val {}\".format( i,maxAction, self.Qtarget[i,maxAction].item() ,  (rewards_mb[i] + gamma * T.max(self.Qs_next_state[i])).item() ) )\n",
        "              self.Qtarget[i,maxAction] = rewards_mb[i] + gamma * T.max(self.Qs_next_state[i])\n",
        "              \n",
        "            target_Qs_batch.append(self.Qtarget[i,maxAction])\n",
        "            #shape (batch_size, num_actions) we want first dim\n",
        "            #We want the loss function to be zero for every action except the maxAction\n",
        "            \n",
        "\n",
        "            \n",
        "            \n",
        "          \n",
        "          #print(self.Qtarget,\"\\n\" ,self.Qs_state,\"\\n\" , (self.Qtarget - self.Qs_state) )\n",
        "          #return;\n",
        "          self.loss       = self.Q_eval.loss(\n",
        "              self.Qtarget, \n",
        "              self.Qs_state\n",
        "          ).to(self.Q_eval.device)\n",
        "          # Optimize the model\n",
        "          self.Q_eval.optim.zero_grad()\n",
        "          self.loss.backward()\n",
        "          self.Q_eval.optim.step()\n",
        "          self.episode_loss.append(self.loss.item())\n",
        "          \n",
        "          \n",
        "          #Get the total reward of the episode\n",
        "          total_reward = np.sum(self.episode_rewards)\n",
        "          total_loss   = np.sum(self.episode_loss)\n",
        "          \n",
        "#           if(self.step % 500 == 0):\n",
        "#             self.save_log(\n",
        "#                 episode,\n",
        "#                 total_reward,\n",
        "#                 self.explore_probability,\n",
        "#                 total_loss / 500 , \n",
        "#                 total_loss, \n",
        "#                 ( time.time() - self.start_time) / 60,\n",
        "#                 self.step,\n",
        "#                 \"Exploration\"  if (self.exploration_vs_explotation) else \"Explotation\",\n",
        "#                 \"Running:\",\n",
        "#                 self.episode_actions\n",
        "#             )\n",
        "            \n",
        "          \n",
        "#           print(self.Qtarget.shape,  self.Qs_state.shape, self.batch_size, states_mb.shape )\n",
        "#           return;\n",
        "\n",
        "        \n",
        "      # Save model every 5 episodes\n",
        "      if episode % target_update == 0:\n",
        "          T.save(self.Q_eval.state_dict(), drive_path+\"/model_deepq_pytorch.ckpt\" )\n",
        "          self.Q_target.load_state_dict(self.Q_eval.state_dict())\n",
        "          print(\"Saving Model and Updating Target\")\n",
        "          \n",
        "          \n",
        "  \n",
        "  def play(self):\n",
        "    state                  = env.reset()\n",
        "    totalScore             = 0\n",
        "    self.decay_step        = 0\n",
        "\n",
        "    for i_episode in range(self.total_episodes):\n",
        "      self.state = env.reset()\n",
        "\n",
        "      self.state, self.stacked_frames  = stack_frames(\n",
        "          self.stacked_frames, \n",
        "          self.state, \n",
        "          is_new_episode=True\n",
        "      )\n",
        "\n",
        "      for t in range(self.max_steps):\n",
        "        if episode_render:\n",
        "          env.render()\n",
        "        #print(self.state)\n",
        "        #action = env.action_space.sample()\n",
        "        action   =  self.Q_eval.forward(T.Tensor(self.state))\n",
        "        action   =  T.argmax(action).item()\n",
        "\n",
        "        self.next_state, reward, done, info = env.step(action)\n",
        "        if done:\n",
        "            self.next_state = np.zeros( (110, 84) , dtype=np.int )\n",
        "            self.next_state, self.stacked_frames = stack_frames(\n",
        "                self.stacked_frames, \n",
        "                self.next_state,\n",
        "                is_new_episode=False\n",
        "            )\n",
        "            print(\"Episode finished after {} timesteps\".format(t+1))\n",
        "            break\n",
        "        else:\n",
        "          self.next_state, stacked_frames = stack_frames(\n",
        "            self.stacked_frames, \n",
        "            self.next_state,\n",
        "            is_new_episode=False\n",
        "          )\n",
        "          self.state = self.next_state"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "6XIAVtz51_AL",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "#Train"
      ]
    },
    {
      "metadata": {
        "id": "E7dQgZoWY3zm",
        "colab_type": "code",
        "outputId": "1ec03c45-201b-419b-8d32-def93ac23ebe",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 892
        }
      },
      "cell_type": "code",
      "source": [
        "agent = Agent(  \n",
        "    stack_size       = stack_size,\n",
        "    state_size       = state_size,\n",
        "    action_size      = action_size,\n",
        "    learning_rate    = learning_rate,\n",
        "    total_episodes   = total_episodes,\n",
        "    max_steps        = max_steps,\n",
        "    batch_size       = batch_size,\n",
        "    explore_start    = explore_start,\n",
        "    explore_stop     = explore_stop,\n",
        "    decay_rate       = decay_rate,\n",
        "    gamma            = gamma,\n",
        "    pretrain_length  = pretrain_length,\n",
        "    memory_size      = memory_size,\n",
        "    training         = training, \n",
        "    episode_render   = episode_render,\n",
        "    possible_actions = possible_actions,\n",
        "    target_update    = target_update\n",
        ")\n",
        "if training:\n",
        "  print(\"Start Training\")\n",
        "  agent.train()\n",
        "else:\n",
        "  print(\"Start Playing\")\n",
        "  agent.play()"
      ],
      "execution_count": 0,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Loadinig Models\n",
            "__________HYPERPARAMETERS__________\n",
            " \n",
            "stack_size         4 \n",
            "state_size         [110, 84, 4] \n",
            "action_size        6 \n",
            "learning_rate      0.00025 \n",
            "total_episodes     20000 \n",
            "max_steps          50000 \n",
            "batch_size         10 \n",
            "explore_start      1.0 \n",
            "explore_stop       0.01 \n",
            "decay_rate         5e-06 \n",
            "gamma              0.9 \n",
            "pretrain_length    10 \n",
            "memory_size        50000 \n",
            "training           True \n",
            "episode_render     False \n",
            "target_update      50 \n",
            "possible_actions\n",
            " [[1 0 0 0 0 0]\n",
            " [0 1 0 0 0 0]\n",
            " [0 0 1 0 0 0]\n",
            " [0 0 0 1 0 0]\n",
            " [0 0 0 0 1 0]\n",
            " [0 0 0 0 0 1]] \n",
            "\n",
            "Start Training\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAABZkAAAFfCAYAAAAszcLlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nzJIkSZIkSZKkQZHRDgAAAAdJREFUof0/uswKQTO5fjgAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f4ba48d3fd0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "Done:\tEpisode:\t0\tTotalReward:\t110.0\tExploreP:\t0.9970\tRunningLoss:\t80404162993.3715\tLoss:\t1608083.2599\tTimeLapse:\t0.4137856483459473\tMin.\tStep:\t50000\tExploration\tAction\t[5, 5, 4, 0, 5, 0, 2, 2, 4, 2, 1, 1, 0, 3, 0, 2, 2, 4, 4, 1, 2, 3, 3, 2, 5, 4, 2, 0, 5, 2, 2, 2, 3, 3, 0, 4, 3, 4, 3, 3, 1, 4, 3, 1, 0, 0, 1, 4, 0, 2, 3, 1, 0, 3, 4, 4, 1, 0, 1, 5, 4, 0, 2, 1, 5, 2, 1, 5, 5, 3, 1, 5, 2, 2, 4, 5, 1, 2, 1, 2, 0, 3, 4, 2, 0, 3, 2, 4, 4, 0, 1, 1, 2, 2, 2, 2, 0, 3, 2, 3, 2, 1, 0, 0, 3, 4, 0, 3, 3, 5, 0, 1, 1, 0, 3, 5, 2, 5, 4, 5, 0, 3, 5, 0, 1, 2, 5, 2, 0, 0, 0, 0, 5, 5, 1, 4, 1, 5, 1, 2, 0, 0, 1, 0, 4, 4, 4, 2, 1, 3, 0, 2, 1, 4, 1, 1, 3, 0, 5, 5, 0, 1, 5, 5, 1, 0, 2, 2, 2, 4, 2, 4, 0, 0, 5, 3, 0, 1, 1, 1, 2, 0, 5, 4, 1, 0, 1, 3, 2, 4, 1, 3, 4, 1, 5, 1, 3, 1, 5, 3, 3, 2, 4, 5, 0, 2, 0, 1, 0, 4, 0, 0, 3, 1, 3, 4, 0, 0, 1, 5, 4, 4, 1, 5, 0, 5, 4, 2, 5, 0, 5, 4, 2, 4, 4, 3, 3, 5, 2, 2, 4, 1, 0, 4, 1, 5, 3, 4, 5, 5, 2, 4, 5, 2, 0, 1, 0, 4, 3, 1, 2, 1, 0, 3, 4, 5, 4, 3, 2, 2, 3, 2, 4, 5, 2, 2, 3, 4, 4, 3, 2, 1, 0, 5, 3, 3, 2, 1, 5, 4, 2, 5, 0, 5, 0, 4, 3, 4, 4, 5, 3, 4, 2, 5, 5, 4, 3, 4, 3, 5, 0, 3, 1, 2, 1, 4, 4, 3, 5, 3, 4, 1, 2, 2, 1, 4, 1, 5, 5, 3, 0, 0, 3, 2, 1, 5, 5, 0, 3, 4, 4, 2, 3, 2, 0, 2, 5, 4, 2, 0, 1, 4, 4, 3, 4, 0, 4, 4, 1, 0, 0, 2, 5, 0, 4, 5, 3, 5, 3, 4, 5, 3, 4, 1, 1, 3, 2, 3, 5, 2, 4, 5, 5, 1, 5, 1, 0, 2, 4, 5, 1, 5, 4, 4, 3, 0, 0, 2, 1, 5, 1, 1, 4, 5, 0, 1, 3, 0, 3, 4, 3, 0, 3, 5, 3, 0, 5, 3, 1, 1, 0, 2, 4, 0, 4, 4, 2, 5, 3, 3, 4, 5, 0, 0, 5, 1, 3, 0, 5, 3, 3, 1, 2, 0, 0, 0, 0, 0, 5, 4, 1, 4, 3, 3, 0, 2, 4, 3, 0, 2, 1, 4, 0, 5, 1, 2, 5, 1, 1, 5, 3, 0, 5, 5, 1, 5, 0, 5, 0, 3, 0, 4, 0, 5, 5, 2, 3, 5, 0, 3, 1, 4, 5, 4, 3, 2, 5, 5, 5, 0, 4, 1, 5, 5, 5, 2, 4, 0, 2, 2, 1, 4, 1, 5, 5, 0, 3, 2, 0, 1, 5, 5, 3, 5, 3, 3, 0, 3, 5, 3, 2, 5, 3, 2, 0, 3, 2, 2, 4, 3, 0, 1, 3, 2, 1, 4, 1, 2, 1, 2, 4, 2, 5, 2, 3, 2, 1, 5, 1, 0, 5, 2, 5, 5, 4, 3, 4, 0, 3, 0, 5, 5, 5, 3, 2, 1, 5, 2, 1, 5, 3, 2, 1, 5, 3, 0, 4, 1, 0, 0, 5, 1, 2, 5, 2, 4, 2, 0, 4, 3, 1, 1, 0, 1, 0, 3, 0, 4]\n",
            "Saving Model and Updating Target\n"
          ],
          "name": "stdout"
        }
      ]
    }
  ]
}